Generalized predictive controller based on RBF neural network for a class of nonlinear system

This paper presents a generalized predictive controller based on RBF neural network (RBF-NN) for a class of nonlinear system with time delay. The procedure of the proposed control system includes two parts: RBF-NN modeling and predictive control algorithm design. The RBF-NN model can predict future outputs of the plant and the predictive value can be amended online, which allows it to employ to complex nonlinear systems. The predictive controller is based on the RBF-NN model and can be used in nonlinear systems with unknown time delay. It can adaptively generate control signals though it is possible that the parameters of the plant are fluctuated or there is noise. The effectiveness of the proposed controller is verified in the simulation of second-order nonlinear systems. Meanwhile, a predictive PID controller is also designed to compare with their performance. It is proven that the generalized predictive controller based on RBF neural network is effective and provided with good adaptation and robustness